Symfony Panther in Production

Panther is Symfony's WebDriver bridge, it drives real Chrome or Firefox from PHP. The API mirrors Symfony's BrowserKit, so transitioning from HttpBrowser to PantherTestCase is mostly mechanical.

For scraping, Panther fills the same role as scrapy-playwright: handle the pages that need JavaScript without leaving the framework.

Install

composer require symfony/panther

Panther needs a browser binary. Locally, Chrome or Chromium. In Docker, install Chromium.

A minimal scrape

<?php
// src/Service/LivePriceScraper.php
namespace App\Service;

use Symfony\Component\Panther\Client;

class LivePriceScraper
{
  public function scrape(string $url): array
  {
  $client = Client::createChromeClient(arguments: [
  '--headless=new',
  '--no-sandbox',
  '--disable-gpu',
  '--window-size=1280,800',
  ]);
  try {
  $client->request('GET', $url);
  $client->waitForVisibility('.price-feed-item');
  $crawler = $client->getCrawler();
  $items = [];
  foreach ($crawler->filter('.price-feed-item') as $node) {
  $sub = new \Symfony\Component\DomCrawler\Crawler($node);
  $items[] = [
  'sku' => $sub->filter('[data-sku]')->attr('data-sku'),
  'price' => $sub->filter('.price')->text(),
  ];
  }
  return $items;
  } finally {
  $client->quit();
  }
  }
}

Key patterns:

• waitForVisibility(selector), block until the element is rendered. Use this anywhere you'd wait_for_selector in Playwright.
• getCrawler(), returns a DomCrawler instance over the current DOM. You can re-use your existing parsing code unchanged.
• quit() in a finally, releases Chrome. Forgetting this leaks browser processes.

Available wait helpers

Method	Use
waitFor(selector)	Wait until selector exists in the DOM
waitForVisibility(selector)	Wait until selector is visible (not hidden)
waitForInvisibility(selector)	Wait until selector disappears
waitForAttributeToContain	Wait for an attribute to contain a value
waitForFunction	Run JS until it returns truthy

Set a longer default timeout:

$client->getWebDriver()->manage()->timeouts()->implicitlyWait(WebDriverTimeouts::SECONDS_30);

Running Panther in Docker

FROM php:8.3-cli

RUN apt-get update && apt-get install -y \
  chromium chromium-driver \
  && rm -rf /var/lib/apt/lists/*

ENV PANTHER_CHROME_BINARY=/usr/bin/chromium
ENV PANTHER_NO_SANDBOX=1

Two environment variables matter:

• PANTHER_CHROME_BINARY, path to Chromium.
• PANTHER_NO_SANDBOX=1, required in containers (Chrome's sandbox needs privileges that containers usually lack).

For headful debugging on Linux, run xvfb-run:

xvfb-run php bin/console scrape:live-prices

Driving Panther from Messenger workers

A Panther-based handler for the live deals endpoint:

#[AsMessageHandler]
final class ScrapeLivePricesHandler
{
  public function __construct(private LivePriceScraper $scraper) {}

  public function __invoke(ScrapeLivePricesMessage $msg): void
  {
  $items = $this->scraper->scrape('https://practice.scrapingcentral.com/deals/live');
  foreach ($items as $item) {
  // persist
  }
  }
}

Each invocation spins up Chrome, scrapes, quits. Chrome is heavy (~200 MB RAM), so don't run too many workers per machine. Tune messenger:consume --limit=N low enough that workers restart frequently, Chrome leaks tend to accumulate over hundreds of sessions.

Reusing Chrome across messages

For higher throughput, reuse the Client across multiple handler invocations:

class LongLivedScraper
{
  private ?Client $client = null;

  public function client(): Client
  {
  if (!$this->client) {
  $this->client = Client::createChromeClient(...);
  }
  return $this->client;
  }

  public function scrape(string $url): array
  {
  $this->client()->request('GET', $url);
  ...
  }
}

Register as a service with shared: true (Symfony default). The same Chrome instance serves many handler runs. Add a periodic quit() to avoid memory drift.

Where Panther falls behind Playwright

Honest comparison:

Feature	Panther	Playwright
Browser support	Chrome, Firefox via WebDriver	Chromium, Firefox, WebKit native
Network interception	Limited	Full request/response mocking
Multi-tab handling	Possible but awkward	First-class
Trace/debug tools	Basic	Playwright Inspector, traces, video
Async / parallel page	Limited	Built-in
Anti-detection	Older WebDriver, Cloudflare detects it more often	Newer, less detected (with some hardening)

For scrapers that need cutting-edge anti-bot evasion, Playwright (via a Python service called from PHP) often wins. For 80% of "wait for JS, extract data" cases, Panther is fine.

A pragmatic hybrid: Symfony does orchestration, persistence, and HTML scraping; a small Python sidecar with Playwright handles the JS-heavy pages. The two communicate via Messenger or a small HTTP API.

Selenium under the hood

Panther speaks WebDriver, the same protocol Selenium uses. Behind Panther is the facebook/webdriver (now php-webdriver/webdriver) library. If you outgrow Panther's API surface, you can use the WebDriver library directly:

$driver = $client->getWebDriver();
$driver->executeScript('return navigator.userAgent;');

Hands-on lab

Against /deals/live on Catalog108 (which streams price updates via WebSocket):

1. Write a Panther scraper that loads the page, waits for .price-feed-item, captures the first 20 prices.
2. Wrap it in a Messenger handler.
3. Schedule it every minute via Symfony Scheduler.
4. Monitor memory across 100 runs, does it stay flat, or does it grow? If it grows, the Chrome reuse pattern is leaking.

Browsers are heavy. Cycle them. Panther in production looks like "use sparingly, restart often."